Quantifying Interactive Cooling Effects of Morphological Parameters and Vegetation-Related Landscape Features during an Extreme Heat Event

: In this study, we apply the ENVI-met model to evaluate the effects of combinations of morphological and vegetation-related landscape features on urban temperatures and thermal comfort. We simulated the thermal conditions of 126 scenarios, varying the aspect ratios of street canyons, vegetation cover and density, surface materials, and orientations toward the prevalent winds under an extreme heat situation. Our results show how the effects of physical and vegetation parameters interact and moderate each other. We also demonstrate how sensitive thermal comfort indices such as temperature and relative humidity are to the built environment parameters during different hours of a day. This study’s ﬁndings highlight the necessity of prioritizing heat mitigation interventions based on the site’s physical characteristics and landscape features and avoiding generic strategies for all types of urban environments.


Introduction
Within cities, urban heat islands and extreme heat events such as heatwaves are some of the most tangible outcomes of climate change that pose direct and indirect effects on citizens' quality of life. Effects range from slight thermal discomfort to extreme heat stress and even heat-related mortalities [1]. According to the United States Center for Disease Control and Prevention, from 2004 to 2018, 702 heat-related deaths occurred on average annually throughout the US [2]. Although the extent of physical and infrastructure loss due to the different levels of vulnerability and exposure may vary spatially and temporally [3], the higher global frequency and intensity of extreme heat events due to climate change [4] is expected to exacerbate the issue even further [5].
In cities worldwide, synergistic interactions between the urban heat island effect and heatwaves can expose a vast number of people to risk [6]. Unfavorable thermal conditions and excessive heat-related mortality, especially among racial minorities and vulnerable groups-such as the case of the Chicago Heat Wave in 1995 illustrated in [7]-are some inevitable outcomes of those situations. Cities, due to the special elements of urbanization (such as imperviousness and land cover) and morphological characteristics [8][9][10], are more vulnerable to the severe impacts of such events. As risk is a function of hazard, exposure, and vulnerability [11], higher risk will be associated with extreme heat events throughout the cities. The impacts of extreme heat exposure can potentially be alleviated through planning strategies for the built environment [12,13]. This notion makes the study of the contributing factors and strategies relevant to the urban planners who are responsible for providing healthy and safe environments for all residents.
The severity, magnitude, and frequency of such extreme heat events in cities are influenced by multiple local-scale factors. Site-specific features such as urban layout, pavement and building materials, and the balance between urban environment, human activities and meteorological conditions can modify microclimates and generate temperature variation within any given city [14,15]. The airflow in urban environments depends mainly on the urban canopy layer characteristics and the temperature differences between ambient air, ground, and surfaces due to radiative and thermal heating [16]. Physical and landscape features of cities such as the geometry of street canyons determined through aspect ratio or sky view factor [10,[17][18][19], street orientations or inflow direction [20][21][22], vegetation (location, distribution, type, and density) [13,[23][24][25], the albedo of surface materials [26][27][28], the presence of water bodies, traffic, and human activities proved to be the most influential in changing temperatures in cities. In addition, the interactions between these parameters and climate-related variables-through altering air and surface temperatures-can also modify thermal conditions in cities [29].
While the effects of aspect ratio, street orientation, vegetation, or surface albedo on temperature variations have been quantified in previous studies, those parameters under different conditions may exhibit different effects [12,30,31] which need to be further scrutinized. For example, tree planting is advocated by many studies as a useful strategy for reducing temperature throughout cities [24,32,33]. However, it appears the cooling effect of trees is largely controlled by background meteorological and morphological conditions. For example, it has been shown that trees and vegetation have a higher cooling impact in sites with taller buildings, higher temperatures, and lower relative humidity [34]. Other results have also shown that trees with high trunk and low canopy density are most effective in improving daytime thermal comfort in deeper canyons [24]. The cooling performance of trees is also dependent on the orientations and surface materials of streets [14,35].
There are many empirical studies that use ground measurements, meteorological data, or remotely sensed imagery to explain the contribution of influential parameters in altering temperature [36][37][38][39]. Yet, due to the complexity of urban areas, high heterogeneity, and high spatial correlation effects, it is difficult both to isolate the causal effects of morphological, meteorological and landscape factors on spatial thermal variation in cities and to understand the contexts in which specific changes to the built environment may exhibit interactive effects that may either amplify or negate improvement to thermal comfort. Physical simulation-based modeling, however, has the capacity to test both the isolated and interactive impacts of vegetation and urban morphological parameters. In addition, the analysis of the temporal outputs from models also allows the examination of the duration of extreme heat conditions, which has important implications for public health and safety.
The purpose of this study is to explore the relative and interactive impacts of potential changes to the built environment during an extreme heat event by analyzing sensitivities of varying parameters that correspond with potential interventions. We do this by simulating a large number of scenarios of an idealized urban street canyon in which the built environment parameters are perturbed in different combinations. Specifically, the following questions will be addressed throughout this paper: RQ.1 Which physical and vegetation parameters cause the largest variations in thermal conditions? RQ.2 What is the effect magnitude of those parameters on changing thermal comfort indices (such as temperature and relative humidity) during different hours of a day? RQ.3 What is the interactive effect of morphological and vegetation-related landscape features in changing thermal comfort levels during an extreme heat event? RQ.4 What is the cooling performance of different vegetation scenarios in various street canyon layouts, and how might they affect the duration of thermal stress and exposure to extreme heat?
Due to the limited capacity of cities for new developments, answers to the above questions would be of exceptional interest to the planners and decision-makers and give new insights into prioritized heat mitigation interventions based on sites' characteristics.

Materials and Methods
For simulating our idealized urban street canyon, we chose the ENVI-met software (version 4.4.1), a popular microclimate computational fluid dynamic (CFD) model, which has been developed to simulate surface-plant-air interactions [40]. ENVI-met has the capacity to provide a typical horizontal resolution of 0.5 to 10 m and a timestep of 1 to 5 s [41]. These relatively high spatial and temporal resolutions facilitate the analysis of complex urban environments. It is also frequently used in scenario analysis for thermal impacts of various kinds of development alternatives [42] and has been tested and validated for heat-related studies in different climates by previous studies [43][44][45]. Since its development in 1998, a keyword search on the Web of Science reveals that ENVI-met appears in 600 academic articles (by the time of writing this paper in 2021).
All of our simulations utilize meteorological forcing conditions that are based on the 1995 Chicago Heat Wave [46]. Variables of the physical simulation domain identified thorough analysis of previous studies (Appendix A Table A1). After developing the scenarios, running the simulations, and obtaining the outputs, relevant indices of thermal comfort were extracted and used for descriptive and inferential statistical analysis. The schematic procedure of our study is shown in Figure 1. Due to the limited capacity of cities for new developments, answers to the above questions would be of exceptional interest to the planners and decision-makers and give new insights into prioritized heat mitigation interventions based on sites' characteristics.

Materials and Methods
For simulating our idealized urban street canyon, we chose the ENVI-met software (version 4.4.1), a popular microclimate computational fluid dynamic (CFD) model, which has been developed to simulate surface-plant-air interactions [40]. ENVI-met has the capacity to provide a typical horizontal resolution of 0.5 to 10 m and a timestep of 1 to 5 s [41]. These relatively high spatial and temporal resolutions facilitate the analysis of complex urban environments. It is also frequently used in scenario analysis for thermal impacts of various kinds of development alternatives [42] and has been tested and validated for heat-related studies in different climates by previous studies [43][44][45]. Since its development in 1998, a keyword search on the Web of Science reveals that ENVI-met appears in 600 academic articles (by the time of writing this paper in 2021).
All of our simulations utilize meteorological forcing conditions that are based on the 1995 Chicago Heat Wave [46]. Variables of the physical simulation domain identified thorough analysis of previous studies (Table A1). After developing the scenarios, running the simulations, and obtaining the outputs, relevant indices of thermal comfort were extracted and used for descriptive and inferential statistical analysis. The schematic procedure of our study is shown in Figure 1.

Model Setup and Scenario Development
ENVI-met for running simulations needs two separate input files: area input file (.IN/.IN3) and model configuration file (.cf/.sim). Each of these two files contains specific sources of data that are essential for the model calculations. In the area input file, mostly physical and spatial parameters such as building shapes and geometry, surface materials, soil profile, vegetation variables, pollutant source, and the receptor location can be identified. In the model configuration file, meteorological parameters such as air temperature, relative humidity, specific humidity, wind speed, wind direction, roughness length, cloudiness level, date, and time of simulation would be modeled. ENVI-met, by matching these two separate files, generates outputs for each hour of simulation [47].
The choice of variables and their values took place based on reviewing the literature, which is reflected in Table A1. In this table, we included the variables, number of scenarios, and main purpose of different studies in which simulation models were used as the primary method. Upon the complexity of dynamic systems, the specification of appropriate initial and boundary conditions could not be overly certain [48]. This in-depth literature review was helpful in the process of parameter calibration and thus made the model

Model Setup and Scenario Development
ENVI-met for running simulations needs two separate input files: area input file (.IN/.IN3) and model configuration file (.cf/.sim). Each of these two files contains specific sources of data that are essential for the model calculations. In the area input file, mostly physical and spatial parameters such as building shapes and geometry, surface materials, soil profile, vegetation variables, pollutant source, and the receptor location can be identified. In the model configuration file, meteorological parameters such as air temperature, relative humidity, specific humidity, wind speed, wind direction, roughness length, cloudiness level, date, and time of simulation would be modeled. ENVI-met, by matching these two separate files, generates outputs for each hour of simulation [47].
The choice of variables and their values took place based on reviewing the literature, which is reflected in Table A1. In this table, we included the variables, number of scenarios, and main purpose of different studies in which simulation models were used as the primary method. Upon the complexity of dynamic systems, the specification of appropriate initial and boundary conditions could not be overly certain [48]. This in-depth literature review was helpful in the process of parameter calibration and thus made the model predictions more reliable. We developed 126 scenarios based on the specified parameters, selected variables, and their different interactions to holistically analyze the solo and interactive effects of variables. Buildings and street canyon geometry (defined by aspect ratio), vegetation type, density and location, street orientation (i.e., toward the prevalent wind direction), and pavement material are among the variables that were analyzed in this study ( Table 1). The interaction of physical and vegetation variables resulted in 108 scenarios which were complemented by 18 scenarios with no vegetation, resulting in overall 126 scenarios. It should be noted here that this study is grounded on microclimate modeling and simulation rather than empirical research. Thus, the scale of analysis is constrained by the limitations of such models. In this study, we focused on an idealized street canyon (the basic component of city structures and urban layouts) which is suited for the purpose of this study and the sensitivity analysis of corresponding parameters. For demonstration purposes, a graphical layout of an idealized street canyon that is simulated in the model has been plotted in Figure 2. In this plot, the H/W ratio (aspect ratio) concept and different wind directions are sketched graphically. predictions more reliable. We developed 126 scenarios based on the specified parameters, selected variables, and their different interactions to holistically analyze the solo and interactive effects of variables. Buildings and street canyon geometry (defined by aspect ratio), vegetation type, density and location, street orientation (i.e., toward the prevalent wind direction), and pavement material are among the variables that were analyzed in this study (Table 1). The interaction of physical and vegetation variables resulted in 108 scenarios which were complemented by 18 scenarios with no vegetation, resulting in overall 126 scenarios. It should be noted here that this study is grounded on microclimate modeling and simulation rather than empirical research. Thus, the scale of analysis is constrained by the limitations of such models. In this study, we focused on an idealized street canyon (the basic component of city structures and urban layouts) which is suited for the purpose of this study and the sensitivity analysis of corresponding parameters. For demonstration purposes, a graphical layout of an idealized street canyon that is simulated in the model has been plotted in Figure 2. In this plot, the H/W ratio (aspect ratio) concept and different wind directions are sketched graphically.  For initializing simulations in addition to variables that are changing through each scenario, in ENVI-met, we need some parameters which are constant for all scenarios. Atmospheric data such as air temperature, relative humidity, and wind speed are among those. For such parameters, we used the meteorological database available for Chicago metropolitan area during the heatwave of July 1995 [46]. Considered values for initial atmospheric boundary layer condition, spatial parameters, and other information regarding location, date, time, and duration of simulations are reflected in Table 1.

ENVI-met Output Post-Processing and Statistical Analysis
The ENVI-met model binary output files (EDX/EDT) were exported to netCDF files, allowing the further analysis of output files in the other platforms. For each scenario, a three-dimensional array (x = 50, y = 50, z = 40) representing the domain conditions at each hour was output for each of 12 timesteps. To avoid the flow inside the canyon being influenced by the domain size, we considered a sufficient number of grids from the borders. All model outputs from the computational fluid dynamics portion of ENVI-met were converted to Universal Thermal Comfort Index (UTCI) using a Python package called "Pythermalcomfort" [49]. Beyond meteorological conditions such as air temperature, thermal comfort needs to take into account various indices of human energy balance such as skin wittedness, clothing, and metabolism [50]. There are several metrics for measuring thermal comfort, such as Predicted Mean Vote (PMV) [51], Perceived Temperature (PT) [52], Outdoor Standard Equivalent Temperature (OUT_SET) [53], Physiologically Equivalent Temperature (PET) [54] and UTCI [55]. UTCI is based on a state-of-the-art thermo-physiological model and has been validated for different climates, seasons, time, and spatial scales and is derived using a non-linear combination of air temperature, relative humidity, wind speed, and mean radiant temperature. The reliability of this indicator for extreme heat situations and comprehensibility of its results by decision-makers (due to the reporting outcomes in centigrade) have made this thermal comfort indicator an appropriate means for comparing the results in this study.
For summarizing and reporting outputs, we used the average values of the response variables (i.e., temperature, relative humidity, wind speed, and mean radiant temperature) throughout the non-building horizontal domain area at the near-surface (pedestrian level) height (<2 m) at each simulation hour t, as is shown in Equation (1) for air temperature.
where x and y denote the position of each grid cell in the horizontal domain at z = 2 m, T refers to air temperature, building footprint refers to cells within the building footprint in the domain, T t is the mean for each simulated output in a timestep which was then used to calculate the UTCI at that timestep, as is shown in Equation (2) below: where T refers to mean air temperature, RH refers to mean relative humidity, WS refers to mean windspeed, and MRT refers to the mean of mean radiant temperature, all at timestep t; f refers to the non-linear function executed using Pythermalcomfort package in Python, and UTCI t refers to UTCI at timestep t.

RQ.1: UTCI Level Sensitivity Analysis
To analyze the sensitivity of UTCI level to different categories of physical and vegetation parameters and examine which parameters cause the largest variations in the UTCI level, we compared the "UTCI level change" of each scenario. The UTCI level change for each scenario indicates the difference between the average UTCI level of the whole canyon at 15:00 (peak time of temperature) and the average UTCI level of the scenario with the highest UTCI level at the same time.

RQ.2: Temporal Analysis
To estimate the effect magnitude of each independent variable temporally, we used a regression-based analysis. Due to the model's long simulation time, we did the simulations for 12 h (10:00 to 22:00) and reported the results of regression analysis starting from 12:00 to 22:00 (the first two hours due to the inaccuracy of the initial simulation hours were cut off from the reports). In Equation (3), we provided the statistical specification of this regression model. Through analyzing the variance of corresponding input parameters and the model response (i.e., thermal comfort indices), we calculated a standardized regression coefficient (SRC) to quantitatively estimate the influence of input parameters on the modeled response. SRC performed as a stable index, especially in linear models and sophisticated global sensitivity methods [56]. SRC values for each timestep were obtained by running the regression model on the standardized form of variables. Standardization happened by implementing the so-called formula for calculating the Z-score (Equation (4)) of the observations of the response variables. SRC can provide good insights on the direction and magnitude of changes in the response with the changes of corresponding input parameters. Ranging from −1 to +1, the sign of SRC explains whether there is a negative or positive effect of the independent variable on the dependent variable, while the zero SRC value demonstrates zero estimated effect of the variable. Moreover, the absolute magnitude of the SRC signifies the parameter's importance, in which higher values indicate the parameter's higher impact on changing the model output while holding the other variables' effect constant. In this study, as all independent variables are categorical, the SRC of each variable is interpreted as the difference in the predicted value of the response (thermal comfort indices) when each category of independent variables switches from the reference point. Here, the reference categories are the shallow canyon, perpendicular orientation toward the wind, the asphalt surface material, and no vegetation for both green type and density variables.
where i refers to i in 1~I, where I = 126 (number of scenarios), t refers to t in 12~T, where T = 24, ε i refers to the error associated with each scenario, ν i,t refers to the error associated with each scenario's timestep, β 0 refers to the intercept, β 1 to β 5 refers to the estimated coefficients of explanatory variables, and Y refers to the dependent variable (e.g., temperature and relative humidity).
where j refers to variable y observations, y refers to the mean of variable y observations, σ refers to the standard deviation of variable y observations, and Z refers to the standard score.

RQ.3: Exploration of Parameter's Interactive Effect
To explore the interactive effect of physical and vegetation parameters and their maximum UTCI reduction potential in different contexts, we estimated the average effect of parameters in UTCI reduction when grouped by other parameters. The reported UTCI reduction in these analyses refers to the difference between the modeled UTCI level of each scenario (which is the average UTCI level of the whole canyon area below the 2-m height at 15:00) and the average UTCI of the whole canyon below the 2-m height in the worst-case scenario (the scenario with the highest average UTCI level) at 15:00.

RQ.4: Scenario Analysis
To compare scenarios temporally in terms of thermal stress level and duration, we plotted the UTCI value distributions (minimum, maximum, average, and quartiles) of scenarios with different green types and densities during simulation hours (12:00 to 22:00) in different canyon types. Moreover, we used the average UTCI values of the grid points of whole canyon areas below 2 m in height to present the hourly UTCI level change in six selected scenarios of vegetation and physical layouts to demonstrate average and extreme scenarios in terms of thermal stress level and duration.

UTCI Level Sensitivity Analysis
In Figure 3, we presented the UTCI level's sensitivity to different categories of parameters. Analyses showed that the UTCI level change is most sensitive to the aspect ratio and changing the depths of the street canyon gives the most control over the UTCI level change. According to Figure 3a, scenarios with very deep street canyons (ratio of 3:1), on average, have the lowest UTCI level and highest potential in UTCI reduction. However, as the highest variance in UTCI level was observed in shallow canyons, we conducted further analyses for the sensitivity of other physical and vegetation parameters to the UTCI level change among this category of street canyons (Figure 3b-e). Comparing the average UTCI level change in canyons with a different orientation toward the prevailing wind showed that the perpendicular orientation has the highest impact in changing and reducing UTCI level compared to the parallel and oblique directions (Figure 3b). The performance of parallel and oblique directions was observed to be almost equal. Among physical parameters, in this study, different categories of surface materials (including asphalt and concrete) did not show a very significant difference to each other in the average UTCI level change and reduction ( Figure 3c). However, the variance of UTCI level was higher among canyons with the asphalt material. Comparing the average UTCI level of shallow canyons with different vegetation types and densities showed that in the case of the presence of vegetation in the canyon, evergreen types of trees in the sparser form of tree planting can have a higher positive impact in reducing the UTCI level (Figure 3d,e). The impact of vegetation in other types of canyons and possible explanations for the previous observation will be discussed in the next sections.

Temporal Analysis
As reflected in Figure 4, the magnitude of effect (i.e., SRC values) of physical parameters on thermal comfort indices such as temperature varies during a day (simulation time). This issue is mainly due to the natural diurnal fluctuations of thermal comfort indices, which are impacted by the urban canopy parameters. The temporal variations of temperature and relative humidity and the effect of built environments on these changes have also been reported in other studies [57,58].
According to our simulations, the impacts of physical factors such as aspect ratio and canyon orientation toward the prevailing wind on changing the near-surface temperature and relative humidity were not constant during different hours of a day. The maximum effects of aspect ratio and canyon orientation on the temperature variation were obtained

Temporal Analysis
As reflected in Figure 4, the magnitude of effect (i.e., SRC values) of physical parameters on thermal comfort indices such as temperature varies during a day (simulation time). This issue is mainly due to the natural diurnal fluctuations of thermal comfort indices, which are impacted by the urban canopy parameters. The temporal variations of temperature and relative humidity and the effect of built environments on these changes have also been reported in other studies [57,58].  Unlike physical parameters, vegetation parameters represented an almost constant effect on changing temperature and relative humidity during different hours of a day. Overall, compared to the physical parameters, vegetation parameters (type and density) showed a lower impact on changing temperature and relative humidity. This result is in line with other studies using CFD models, which investigated the role of greenery quantity and type and reported the slight ambient cooling effects of high-density vegetation [59][60][61]. The blocking effect of vegetation, especially when planted with high densities on wind speed and outgoing longwave radiation, has been identified as the main reason for its lower impact on thermal comfort enhancement at the street canyon scale [62,63].
However, other studies have shown that vegetation can minimize the solar radiant absorption and lower surface and air temperature through shading and evapotranspiration [64,65]. Thus, when vegetation is planned to be implemented as a heat mitigation strategy, decision-makers might need to consider different approaches to maximize their positive impact and reduce the blocking effect. For example, promoting other forms of vegetation cover, such as building-facade and rooftop greening, using particular types of trees, or spatial considerations of new greenery, can be among those strategies. According to our simulations, the impacts of physical factors such as aspect ratio and canyon orientation toward the prevailing wind on changing the near-surface temperature and relative humidity were not constant during different hours of a day. The maximum effects of aspect ratio and canyon orientation on the temperature variation were obtained at 15:00 (maximum temperature hour), while minimum effects were observed at the early hours of the day. The negative SRC values of the aspect ratio in Figure 4 shows that changing to deeper canyons (higher aspect ratios) can lower near-surface temperatures during the daytime. The positive value of canyon orientation in the same plot shows that changing the canyon orientation toward the prevailing wind from the perpendicular to parallel and oblique would result in an increase in the canyon near-surface temperature during the daytime. Moreover, in this study, changing the canyon pavement's surface material from asphalt to concrete resulted in a slight decrease in near-surface temperature during the daytime; however, this effect faded through the later hours of simulation time. To further investigate the significance of each variable on changing temperature, in Table 2, we reported the regression p-values at each timestep. According to this table, there was a statistically significant relationship between the physical parameters and temperature variations in most timesteps.
As it was expected, the effect magnitude of physical parameters on changing relative humidity shows an opposite trend, yet the maximum effects still were obtained at 15:00. Extracting the hourly SRC value of physical parameters for the wind speed and mean radiant temperature (the other two important indices of thermal comfort) showed a constant value of SRC, which implies the time independence of physical parameters' effects on these two parameters.
Unlike physical parameters, vegetation parameters represented an almost constant effect on changing temperature and relative humidity during different hours of a day. Overall, compared to the physical parameters, vegetation parameters (type and density) showed a lower impact on changing temperature and relative humidity. This result is in line with other studies using CFD models, which investigated the role of greenery quantity and type and reported the slight ambient cooling effects of high-density vegetation [59][60][61]. The blocking effect of vegetation, especially when planted with high densities on wind speed and outgoing longwave radiation, has been identified as the main reason for its lower impact on thermal comfort enhancement at the street canyon scale [62,63]. However, other studies have shown that vegetation can minimize the solar radiant absorption and lower surface and air temperature through shading and evapotranspiration [64,65]. Thus, when vegetation is planned to be implemented as a heat mitigation strategy, decision-makers might need to consider different approaches to maximize their positive impact and reduce the blocking effect. For example, promoting other forms of vegetation cover, such as building-facade and rooftop greening, using particular types of trees, or spatial considerations of new greenery, can be among those strategies.

Parameter Interactive Effects
According to Figure 5a, in the shallow canyon, the presence of high-density vegetation did not considerably reduce the UTCI level, while in deeper canyons, a higher density of vegetation resulted in a significant UTCI level reduction. In Figure 5b, we examined the potential impact of two different tree types (deciduous and evergreen) with different densities. The result showed the sparser density of the deciduous and higher density of evergreen tree types are more effective in reducing the average UTCI level in canyons. Figure 5c represents the relative effect of different orientations of street canyons toward the prevalent wind with respect to the aspect ratio. Overall, the perpendicular orientation appeared to be more effective in reducing UTCI level; however, in shallow canyons, this orientation was significantly more favorable for UTCI reduction rather than oblique and parallel directions.

Parameter Interactive Effects
According to Figure 5a, in the shallow canyon, the presence of high-density ve tion did not considerably reduce the UTCI level, while in deeper canyons, a higher de of vegetation resulted in a significant UTCI level reduction. In Figure 5b, we examine potential impact of two different tree types (deciduous and evergreen) with different sities. The result showed the sparser density of the deciduous and higher density of green tree types are more effective in reducing the average UTCI level in canyons. F 5c represents the relative effect of different orientations of street canyons toward the p alent wind with respect to the aspect ratio. Overall, the perpendicular orientation peared to be more effective in reducing UTCI level; however, in shallow canyons orientation was significantly more favorable for UTCI reduction rather than oblique parallel directions.

Scenario Analysis
To further examine vegetation parameters' contextual impacts on street cany thermal condition and the duration of heat stress, in Figures 6-8 we plotted the U distributions (minimum, maximum, average, and quartiles) of scenarios with diff green types and densities (Table 3) during simulation hours (12:00 to 22:00) in diff canyon types. Each boxplot corresponds to the UTCI range of a single scenario ove simulation hours. According to these plots, in most scenarios in shallow canyons UTCI reached the extreme heat-stress level (with an average duration of 2 h and 15 However, in deep and very deep canyons, none of the scenarios' peaks reached th treme heat stress level. The average duration of the very strong heat stress level for low canyon scenarios was observed to be 5 h and 16 min, and for deep and very (a) the interaction between aspect ratio and green density (b) the interaction between green type and green density (c) the interaction between aspect ratio and canyon orientation.

Scenario Analysis
To further examine vegetation parameters' contextual impacts on street canyons' thermal condition and the duration of heat stress, in Figures 6-8 we plotted the UTCI distributions (minimum, maximum, average, and quartiles) of scenarios with different green types and densities (Table 3) during simulation hours (12:00 to 22:00) in different canyon types. Each boxplot corresponds to the UTCI range of a single scenario over the simulation hours. According to these plots, in most scenarios in shallow canyons, the UTCI reached the extreme heat-stress level (with an average duration of 2 h and 15 min). However, in deep and very deep canyons, none of the scenarios' peaks reached the extreme heat stress level. The average duration of the very strong heat stress level for shallow canyon scenarios was observed to be 5 h and 16 min, and for deep and very deep canyons, it was observed to be 4 h (Figure 9). Moreover, the UTCI level fluctuation was more significant in shallow canyons; this can be attributed to the greater exposure of such canyons to solar radiation and more absorption and reflection. In contrast, in deeper canyons, the enclosure of the street area by tall buildings can hamper both the ventilation and radiant reflection and result in a more constant thermal situation and a lower difference between the minimum and maximum UTCI levels. Comparing different vegetation scenarios in different canyon types also shows that in shallow canyons, the presence of vegetation (especially in higher density and with deciduous green type) raised the average UTCI level, while in deeper canyons, the presence of a higher density of evergreen types of trees lowered both the average UTCI level and the duration of very strong heat stress level.                   In Figure 9, we presented the UTCI level change in six selected scenarios (their characteristics can be found in Table 3) during simulation hours. Those scenarios were chosen based on analyzing the hourly distribution of the UTCI level data points of all scenarios. The fluctuations and a spike in UTCI level at 15:00 in shallow canyons are also evident in this plot (S2_A and S6_A). The 8.5 • C difference between the maximum temperature of worst and best-case scenarios (S2_A and S6_C) in terms of UTCI level shows the impact of both physical (aspect ratio) and vegetation (type and density) parameters in changing the hourly thermal comfort level. Figure 10 also shows the spatial distribution of simulated air temperature at 15:00 in these scenarios. Red and yellow colors in these plots demonstrate higher air temperatures, whereas the green and blue colors represent lower air temperatures. The overall temperature level of different scenarios is in line with the UTCI performance, where green scenarios in shallower canyons exhibit higher temperatures and in deeper canyons demonstrate lower temperatures. However, the distribution of air temperature in all those plots follows a similar pattern, in which shaded areas (including inside canyons) have lower temperatures compared to the unshaded areas (at 15:00). these scenarios. Red and yellow colors in these plots demonstrate higher air temperatures, whereas the green and blue colors represent lower air temperatures. The overall temperature level of different scenarios is in line with the UTCI performance, where green scenarios in shallower canyons exhibit higher temperatures and in deeper canyons demonstrate lower temperatures. However, the distribution of air temperature in all those plots follows a similar pattern, in which shaded areas (including inside canyons) have lower temperatures compared to the unshaded areas (at 15:00).

Limitations
The results of this study were confined by limitations related to the applied microclimate model. Although we attempted to increase the robustness of our results by considering multiple scenarios and changing the input values, like other numerical simulation models, the ENVI-met model is sensitive to initial boundary conditions and input parameters. Therefore, in future studies, by applying a variety of meteorological conditions, we can analyze both the model and parameter sensitivity under various thermal situations to reduce the model uncertainty and expand the applicability of its results for various climatic conditions. The study is also completely based on numerical simulation, which allows for controlled testing of the interactive and isolated effects of various parameters but would be complemented by additional empirical data collection in diverse urban canyon environments.
In addition, in this study, the higher-level effects of parameters such as vegetation have not been considered. For example, the influence of vegetation on the local air quality

Limitations
The results of this study were confined by limitations related to the applied microclimate model. Although we attempted to increase the robustness of our results by considering multiple scenarios and changing the input values, like other numerical simulation models, the ENVI-met model is sensitive to initial boundary conditions and input parameters. Therefore, in future studies, by applying a variety of meteorological conditions, we can analyze both the model and parameter sensitivity under various thermal situations to reduce the model uncertainty and expand the applicability of its results for various climatic conditions. The study is also completely based on numerical simulation, which allows for controlled testing of the interactive and isolated effects of various parameters but would be complemented by additional empirical data collection in diverse urban canyon environments.
In addition, in this study, the higher-level effects of parameters such as vegetation have not been considered. For example, the influence of vegetation on the local air quality might change the radiation balance and hence the thermal condition. Another important fact that should be mentioned here is the distinction between the microscale and mesoscale effect of the mitigation strategies such as vegetation. While the current study mainly focused on the microclimate in an idealized street canyon, the vegetation effect might have a different range on a broader scale. For example, comparing microscale and mesoscale simulations' results demonstrates that a more intensive green scenario has a more accentuated effect on thermal comfort level at the mesoscale rather than the microscale [66]. Thus, estimating the trade-offs between the microscale and mesoscale impact of mitigation strategies for obtaining a more vivid picture of their actual effects seems important and necessary.

Conclusions
This paper analyzed the isolated and interactive effects of physical and vegetation parameters on thermal conditions in an idealized street canyon under an extreme heat situation. We conducted a sensitivity analysis of varying parameters that correspond with potential interventions such as greening and surface materials in various canyon geometries and orientations. Our results confirmed that the aspect ratio has a pivotal role in changing the UTCI level and can control other variables' impact, such as vegetation. Analyzing the hourly UTCI level of scenarios with different vegetation types and densities demonstrated that the duration and magnitude of heat stress is largely controlled by the morphological context. We also analyzed the daily patterns of physical and vegetation parameters' impacts on thermal comfort indices (temperature and relative humidity) and noticed the magnitude of the aspect ratio and canyon orientation's effects fluctuate over a day, being magnified during the hottest hours of the day.
Comparing the hourly UTCI level of several scenarios and their characteristics proved that no fixed strategy exists for providing better thermal conditions, and the combination of different parameters can create various thermal situations. Therefore, special attention should be paid to the morphological and landscape features of urban areas when choosing heat mitigation strategies. Mitigating extreme heat and reducing the duration of exposure in cities would have critical implications for public health and safety. Regarding the limited capacity of cities for new developments, the results of such studies can be useful for urban planners and decision-makers to prioritize interventions based on sites' characteristics. Such results can help avoid generic and inefficient heat mitigation strategies, which might even worsen the situation during an extreme heat event.

Conflicts of Interest:
The authors declare no conflict of interest.